Anthropic's AI Just Got Cheaper and More Private

Anthropic just launched new versions of its advanced AI models, Fable 5.1 and Mythos 5.1, with two big upgrades: they’re cheaper to use and easier to run privately. Fable 5.1 is now available to anyone through cloud platforms or APIs, while Mythos 5.1 is limited to cybersecurity and life sciences partners. Both models promise better performance, but the real headline is privacy—companies can now run these AIs on their own computers without sending data to Anthropic, a feature called zero data retention.

One of the biggest changes is the upcoming rollout of Enterprise Frontier Safeguards, a high-privacy service that lets businesses control how the AI is monitored for misuse. Previously, Fable couldn’t run on private infrastructure due to security concerns, but now it can—with safeguards still in place. Anthropic also claims it has never trained on enterprise data without permission, and won’t start now. This should ease worries about sensitive information being used to train future models.

Alongside the new models, Anthropic shared three scientific discoveries the AI helped uncover before release, including a GPU optimization trick and a high-resolution map of Venus. The company also released detailed system cards explaining the models’ strengths and weaknesses. Mythos 5.1 is slightly more prone to misbehavior than some other models, but it’s better at following rules and less likely to hallucinate fake tasks.

For businesses, this means more control over AI tools, lower costs, and the ability to keep data in-house. For everyone else, it’s a sign that AI is getting faster, cheaper, and more customizable—even if the tech giants keep the most advanced versions for themselves.